You are on page 1of 16

REVIEW OF CLASSICAL DESIGN METHODS AS APPLIED TO

ALUMINUM BILLET HEATING WITH INDUCTION COILS

Mark William Kennedy1, Shahid Akhtar1, Jon Arne Bakken1, Ragnhild E. Aune1,2
1
Department of Materials Science and Engineering, Norwegian University of Science and
Technology, N-7491 Trondheim,
NORWAY
2
Department of Materials Science and Engineering, Royal Institute of Technology, 100 44
Stockholm,
SWEDEN
Communicating author: ragnhild.aune@ntnu.no

Keywords: Induction, heating, billet, short coil

Abstract

In the present study classical induction design tools are applied to the problem of heating non-
magnetic metal billets, using 50 Hz AC. As an example of great practical industrial interest, the
induction heating of aluminum billets is addressed specifically. The predicted work piece power
is compared with the measured work piece power for a long and a short coil, using well
established methods, such as those of Burch and Davis, introduced in 1926/28, Dwight and
Bagai in 1935, Baker in 1944/57, Vaughan and Williamson in 1945, and by Tudbury in 1960. A
calculation methodology based on a combination of the available tools is also introduced and
discussed. The method has proven to give an error of <10% of the actual work piece power. An
equation for Tudbury's work piece shortness correction factor is disclosed for the first time.

Introduction

Induction heating of metal billets is a practical problem, with a geometry suitable for solution
using analytical methods. This paper will examine solutions related specifically to round work
pieces in round coils, although the methods presented can be extended to other regular shapes of
work pieces and coils[1]. The methods can also be extended to ferromagnetic materials with
slight changes[2-3]. It is assumed that the coil and work piece have a consistent diameter over
their entire length and that the work piece is at least as long as the coil. Correction factors can be
derived for the cases where the work piece is shorter than the coil.

The study of induction heating by coils began in the late 19th century with theoreticians like
Heaviside[4]. The work of Burch and Davis[5-7] in the 1920’s greatly improved the theoretical
understanding of induction applied to metallurgy. Many investigators in the 1930’s-1950’s:
Dwight and Bagai[8], Baker[1-2, 9], Vaughan and Williamson[3, 10] and others[11-12] added to
our understanding with clearer mathematics, solutions for specific applications, semi-empirical
modifications and practical design tools. Much of the accumulated knowledge is available in
books, such as Tudbury’s practical text or Simpson’s engineering guide from 1960[13-14].
Perhaps the best description and derivation of the theoretical and semi-empirical solutions can be
found in Davies[15].
In coreless induction heating, a work piece is placed within an induction coil as shown in Figure
1. Both the work piece and coil can be of arbitrary shape and length. A cyclically varying
voltage is impressed upon the coil, which creates a time varying current in the coil and a strong
magnetic field along the axis. This varying magnetic field interacts with the work piece to
induce circular voltage gradients and “eddy currents,” which circulate around the axis of the coil,
but in a direction opposite the coil’s current, as shown in Figure 1. The circulation of the eddy
currents, J (A/m2) and the work piece’s natural resistivity, ρ (ohm·m), combine to generate heat,
P (W/m3). The relationship between the coil current, the magnetic field strength, the induced
voltages and currents, is complex and will be discussed in some detail. Due to the relative
simplicity we will begin by discussing an empty or “air core” coil.

Figure 1. 10 turn induction coil, with billet slightly longer than coil

Derivation of Coil Impedance and Power Formulae

Let us begin by considering the coil shown in Figure 1. It has a length (lc) that can be measured
in two ways: taking the centre-centre distance on the sides of the leads or top-bottom of the coil
turns opposite the leads (as shown in Figure 1).

Imagine that the coil was constructed from a solid copper tube of length (lc). Let the thickness of
the copper (tc) be very small in comparison with the diameter of the cylinder (Dc). (Nc-1)
grooves of thickness (sc) are cut in a helical spiral along the whole length. This forms a long coil
or solenoid, with a geometry approaching that of a theoretical “current sheet” with Nc turns as
shown in Figure 2.

The coil pitch is equal to:


pc = dc + sc (1)

The coil space factor is the fraction of the side of the coil occupied by copper:

kr = dc Nc / lc = dc Nc / [(Nc -1) sc + dcNc] or for large Nc, kr ~dc/pc (2)


Figure 2. 10 turn helical “current sheet” coil or solenoid

The length of the “wire” in the coil, at the average current depth is:

ℓwire = π (Dc + δc) Nc (3)

The height of each “wire” is:


dc = lc kr / Nc (4)

We now apply an alternating voltage (AC), at a frequency, which has an electromagnetic


penetration depth (δc) of much less than the thickness of the conductor in the radial dimension,
(δc<<tc). Where the penetration depth is defined as the depth, which if filled with a
homogeneous current (DC current) would yield the same resistance as the AC current
distribution over the whole thickness of the conductor. At high frequency, current in a solenoid
is biased towards the surface of the wire facing the interior of the coil and attenuates in an
exponential manner with each penetration depth into the conductor.

The penetration depth can be calculated from the following commonly accepted formula[16-17]:

δc = (2 ρc / (ω μr μo))0.5 = (ρc / (π μr μo f ))0.5 (5)

Where: μo is the magnetic permeability of free space, (4 π 10-7 H/m) and μr is the relative
permeability for copper and aluminum, which is equal to 1.

The angular frequency is:


ω = 2 π f (radians per second) (6)

Where: f is the frequency (Hz).

A modified penetration depth can be calculated accounting for the presence of the air gaps as
derived by Howe[18] and with demonstrated accuracy according to the data of Vaughan and
Williamson [10], using the kr factor from Equation (2):

δc = (2 ρc / (kr ω μr μo))0.5 = (ρc / (kr π μr μo f ))0.5 (7)

The alternating current resistance of the coil can then be found using the simple formula:

Rc = ρc ℓwire / acurrent (8)


Where the effective area of the current path for alternating current is:

acurrent = dc δc = δc lc kr / Nc (9)

The resistivity of the copper is a function of temperature and composition as shown in Figure 3.
Phosphorous deoxidized copper (with up to 0.04 wt% residual P) will have an actual
conductivity of about 80% IACS (International Annealed Copper Standard of 1913[19]). 100%
IACS is equivalent to a resistivity of 1.7241E-8 ohm·m at 20oC, with copper having a
temperature coefficient of 0.00393[19-21].

Actual coil copper resistivity can be estimated by:

ρc = 1.7241E-8 ohm·m (1+0.00393 (Tc-20oC)) 100 / %IACS (10)

We can now estimate using equations (1-10) the AC resistance of the coil to be:

Rc = π (Dc + δc) Nc ρc / (δc lc / Nc) = π (Dc + δc) Nc2 ρc / (δc lc kr) (11)

Figure 3. Approximate effect of impurities on the resistivity of copper[20]

Power lost from the coil to the cooling water can then be calculated from P=Ic2Rc, where Ic is the
root mean square (R.M.S.) value:

Pc = Ic2 π (Dc + δc) Nc2 ρc / (δc lc kr) (12)

If the electromagnetic penetration depth exceeds more than half the actual conductor thickness,
as might occur at low frequency (tc>δc>0.5tc), some small error will be encountered with
equations (11-12). If the penetration depth is significantly greater than the thickness (δc>tc), the
resistance can be accurately estimated using the total conducting area (equivalent to the DC
resistance in this case) in Equation (8). Care must be taken to account for the loss of conducting
area due to the cooling channel. In other cases, either an “exact” theoretical solution[18, 22-28]
or empirical data[29] must be used. There is a tendency for the “exact” solutions to be in error
for very short coils (lc/Dc<1), or coils with large spaces between the turns (kr<0.7). It is best if
the coil can be designed such that (δc<0.5tc), for low resistive losses and accurate design.

Figure 4. Examples of conductor shapes for mains frequencies (50-60 Hz), medium (500-5000
Hz), and high frequency furnaces, each with the recommended wall thickness of 2δ[30]

It is standard practice in induction furnace coil design to assume that the coil is magnetically
thick (δc<0.5tc). It is then assumed that the wire of the coil has an internal reactance equal to the
resistance, which is the same as saying that the wire has an internal power factor of 0.707:

X c = Rc (13)

This is a dubious assumption except for very high ratios of thickness to penetration depth
(δc<0.25tc), but causes little error in the total estimation of reactive power. Errors are smallest at
high frequency with thin conductors as shown in Figure 4.

Magnetic Field, Inductance and Impedance for Short and Long “Air Core” Coils

If we imagine that our coil is merely a section of length lc of an infinitely long coil, the magnetic
flux density within the coil is independent of position in the coil. The magnetic flux density can
then be estimated using either Amperè’s or the Biot-Savart law to be:

B∞ = μo Nc Ic / lc (14)

This assumes negligible flux density outside of the coil (no reluctance in the external magnetic
circuit), ignores end effects (there are no ends to an infinite coil) and assumes complete linking
of all the flux with all the turns of the coil.

The inductance of the empty long coil (“air core” inductor) is then determined from the number
of flux linkages per unit current:

L∞ = Ac μo N2c / lc = Ac Nc B∞ / Ic (15)

Where Ac is the area of the coil:

Ac = π (Dc+δc)2 / 4 (16)

For round tubes or if δc≥ tc, the diameter at the centre line of the conductor should be used.
Induction furnace coils are nearly always “short coils”, lc/(Dc+δc)<8, and are often very short
lc/(Dc+δc)<1. There are fundamental differences between long and short coils. In the case of
short coils, the assumptions of uniform internal and negligible external flux density are no longer
correct. The short coil flux density is now a function of both length and radius [Bshort(l, rc)] and
there now exists an external magnetic flux density near the coil, with a finite reluctance. The
approximation that all the flux, links all the turns is also questionable.

In the case of short “air core” coils, exact solutions have been calculated for the “end effect” of a
“current sheet” and a correction factor (kN) tabulated by Nagaoka[31-32]. The short coil
inductance can then be calculated by:

L0 = kN Ac μo N2c / lc = kN Ac Nc B∞ / Ic (17)

Equation (17) is the equivalent to saying that the actual flux density (integrated by length and
radius) of the short coil in the axial direction is:

B0 = kN B∞ = kN μo H∞ = kN μo Nc Ic / lc (18)

The Nagaoka factor has been plotted in Figure 5. Alternatively, an empirical equation like that
of Wheeler[33], can be used to estimate coil inductance for a short coil. Wheeler’s equation has
been reformulated by Knight[34] to give the Nagaoka coefficient directly:

kN = 1/[1 + 0.4502 (Dc + δc) / lc)] (20)

0.9

0.8

0.7
2
kNagaoka , kNagaoka

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
D c /l c
kNagaoka kNagaoka^2

Figure 5. Nagaoka’s short coil correction factor

If round tubing is used, it is necessary to apply a round wire correction such as Rosa’s[32] to the
calculated inductance.
Once the inductance of the short coil has been obtained, it is then simple to estimate the
inductive reactance (X0), the total coil impedance (Z0) and voltage drop (V0) for the “air core”
coil:
X0 = ω L0 = 2 π f L0 (21)

Z0 = (Rc2 + X02)0.5 (22)

V0 = I0 Z0 (23)

Impact of the Work Piece on the Magnetic Field

In the presence of a work piece, we can see from Figure 1, that the total magnetic flux can be
divided into 3 parts: flux which links only with the coil (Φc), flux in the air gap (Φg) and the net
flux that links the billet (Φw). These fluxes in turn produce equivalent inductances and reactive
impedances: Xc, Xg and Xw’, as indicated in the simplified series circuit diagram, Figure 6.

Several things happen to the magnetic field when a work piece is placed into the coil:
1. The “aperture” of the coil is effectively reduced, making the effective diameter to length
ratio look smaller, changing the magnitude of the short coil correction factor kN and the
flux density in the air gap along the axial direction of the coil.
2. Space that was formerly air is occupied by the work piece, reducing the total air gap flux.
3. Eddy currents in the billet produce magnetic fields, which oppose the magnetic field of
the coil, reducing the net flux in the space occupied by the billet.

Vaughan and Williamson proposed an empirical modification to the short coil correction factor
based on the volumetric fraction of the air-gap occupied by the work piece. Their proposed
correction factor is applied to the Nagaoka coefficient to produce an “effective Nagaoka
coefficient” for a short coil containing a work piece:

kN* = kN (1 - [Dw / Dc]2) + [Dw / Dc]2 (24)

To be consistent with the previous formulae:

kN* = kN (1 - [Dw / (Dc + δc)]2) + [Dw / (Dc + δc)]2 (25)

The effective magnetic flux density in the air gap of a short coil containing a work piece can then
be calculated by:

B0* = kN* μo H∞ = kN* μo Nc Ic / lc (26)

It is assumed that this magnetic flux density is constant over the width of the air gap, and that it
is this flux density, which induces the eddy currents in the work piece.

The contribution of the air gap to the total inductive impedance of the circuit is:

Xg = ω Lg = 2 π f Ag Nc B0* / Ic = kN* μo Nc2 π2 (Dc2 – Dw2) / 2 lc (27)


Derivation of Work Piece Impedance and Power Formulae

It is difficult to make electrical measurement in the work piece. The effect of the work piece on
the circuit is normally inferred from electrical changes in the coil, with and without a work piece.
As a first approximation it can be assumed that the resistance and reactance of the coil are not
affected by the work piece and all measured electrical changes can be attributed to the work
piece and changed air gap.

One approach to the mathematics is to consider the magnetic and electrical circuits to be a series
circuit as shown in Figure 6, for a “long coil”. The picture is similar, but more complex for a
“short coil” if the reluctance of the external magnetic circuit is to be accounted for explicitly[1].

Where RW’ and XW’ are the work piece


resistance and reactance referred to the
coil. See Equations (39-40).

Figure 6. Series equivalent circuit for an induction furnace coil and work piece

When alternating current is applied, the coil and work piece become inductively coupled in a
similar way to the windings of a transformer over the length shared by the coil and work piece.
The magnetic flux density produced in the air gap by the coil, is the same flux density, which
induces eddy currents at the surface of the work piece:

Bw = B0* = μo μr Nw Iw / lw = kN* μo Nc Ic / lc (28)

If lw ≥ lc take lw = lc, i.e. it is assumed that only the length of work piece within the coil interacts
with the magnetic field of the coil. In the following derivations we take lw ≥ lc and substitute lc in
the place of lw.

Nw Iw = kN* Nc Ic (lc / lc) = kN* Nc Ic (29)

The eddy currents induced in the billet are assumed to make a single turn (Nw=1), while the
current in the coil makes Nc turns. It is standard practice to refer conditions in the work piece
back to the coil (where we measure the voltage, current, resistance and impedance of the total
circuit) via the standard transformer relationship:

Iw = kN* Nc Ic / 1 = kN* Nc Ic (30)


By analogy to Equation (11), with Nw=1, kr=1 (as the work piece has no “air gaps”) and the
average diameter is reduced by the penetration depth:

Rw = π (Dw - δw) Nw ρw / (δw lc / Nw) = π (Dw - δw) ρw / (δw lc ) (31)

Where δw can be calculated using Equation (5) substituting work piece (w) for coil (c) values.
If the work piece is shorter than the coil, the calculated resistance must be corrected by
multiplying by the ratio (lw / lc).

Using Equations (30-31), the power in the work piece is then:

Pw = Iw2 Rw = kN*2 (Ic Nc)2 Rw = kN*2 (IcNc)2 π (Dw - δw) ρw / (δw lc) (32)

Equation (32) is sufficiently accurate (about 1% error) provided that ξw>3:

ξw = Dw / (δw √2) (33)

ξw as mentioned above is a very important dimensionless parameter in induction heating theory,


and can be found in many different forms in various publications.

The actual resistance in the work piece can be “reflected” back onto the coil via:

Rw’ = Rw Nc2 (34)

and added to the coil resistance, to give the total circuit resistance as indicated in Figure 6:

Rt = Rw’ + Rc (35)

Higher mathematical precision requires the exact solution to the field equations using complex
numbers and involves use of (computationally challenging) Bessel functions[4, 6, 8]:

Rw’ = [π Nc2 Dw ρw / (δw lc)] √2 (berξw ber’ξw + beiξw bei’ξw) / [ber2(ξw) +bei2(ξw)] (36)

Substituting Equation (33) in (36):

Rw’ = (√2 π Nc2 ρw ξw / lc) √2 (berξw ber’ξw + beiξw bei’ξw) / [ber2(ξw) +bei2(ξw)] (37)

Defining φ(ξw):

φ(ξw) = √2 (berξw ber’ξw + beiξw bei’ξw) / [ber2(ξw) +bei2(ξw)] (38)

Substituting φ(ξw) in (37):

Rw’ = √2 π Nc2 ρw ξw φ(ξw) / lc (39)

Similar solutions exist for the inductive reactance of the work piece[4, 6, 8]:

Xw’ = √2 π Nc2 ρw ξw ψ(ξw) / lc (40)


Where ψ(ξw) is defined as:

ψ(ξw) = √2 (berξw bei’ξw - beiξw ber’ξw) / [ber2(ξw) +bei2(ξw)] (41)

Substituting Equations (39) and (34) into Equation (32) for the power developed in the work
piece:

Pw = kN*2 √2 π (Ic Nc)2 ρw ξw φ(ξw) / lc (42)

And similarly for the reactive power of the work piece:

Qw= kN*2 √2 π (Ic Nc)2 ρw ξw ψ(ξw) / lc (43)

φ(ξw) and ψ(ξw) are plotted in Figure 7 versus ξw for use with Equations (42-43). Baker presents
the equations in slightly different format, using his functions P and Q, which are similar
solutions for the real and imaginary parts of the Bessel functions[1]. Excellent derivations of
Baker’s P and Q can be found in the literature[14-15].

The power factor of the work piece can be calculated by:

P.F. = φ(ξw) / [φ(ξw)2 + ψ(ξw)2]0.5 (44)

and becomes close to constant for ξw>3, converging slowly towards 0.707 or √2/2, for very high
ξw.

1.2
φ(ξw), ψ(ξw), Power Factor, unitless

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
ξw, unitless

φ(ξw) ψ(ξw) Power factor

Figure 7. Resistance and Reactive factors for use with Equations (42-43).
Work Piece Shortness Correction Factor

As the length of a coil increases for a fixed diameter, the Nagaoka coefficient and the modified
coefficients kN* and kN*2 approach unity and the ratio of the coil diameter to length (Dc/lc)
approaches zero at the limit. As the diameter of the work piece is reduced with a constant coil
diameter, the coil increasingly looks like an “air core” coil, and at the limit of Dw/Dc going to
zero, kN*2 = kN2.

The modified Nagaoka coefficient squared (kN*2) of Vaughan and Williamson[10], is plotted in
Figure 8. If this figure is compared with the original graph in the book by Tudbury[13], it is
clear that they are identical and that the equation fits the logical tests given above. See Figure 5
for the comparison of kN2 with the intercepts on Figure 7 (limit of the “air core”). Tudbury’s
original book must be referenced, as later reproductions are not drawn accurately[35].

1
Modified Nagaoka Coefficient 2, kN*2

0.9 Ratio of Coil


0.8 Diameter to
Length, Dc/l c
0.7
0.6
0
0.5 0.2
0.4 0.4
0.6
0.3
0.8
0.2 1
0.1
0
0 0.2 0.4 0.6 0.8 1

Ratio of load to coil diameter, Dw/Dc

Figure 9. Tudbury’s Work Piece Shortness Factor and Vaughan and Williamson’s modified
Nagaoka Coefficient2[10, 13]
Results and Discussion

Experiments were conducted using a number of aluminum billets of different sizes and
compositions, using two different coils as indicated in Table I. Billet electrical conductivities
were estimated by simultaneously minimizing the error from Equation (42) and Baker’s method,
while ensuring the conductivity fell within the range of published specifications for each alloy.
Coils were constructed using thin wall tubing for convenience (as power efficiency was not an
issue), where the AC resistance was identical with the DC resistance (δc>dc>tc). This eliminated
the need to have capacitor banks and reduced the impact of harmonics on the measurements.
Copper tubing conductivity was ~80% IACS. Measurements were taken with a Fluke 43B Power
Quality Analyzer, equipped with a calibrated Fluke i1000s current probe with an accuracy of 1%.

Work pieces were cooled to below room temperature (~10oC) and then readings were taken
before they heated beyond 30oC, such that 20oC was a typical temperature to evaluate the
resistivity. Ideally the work pieces would have been drilled through, cooled with water to
achieve isothermal conditions and the heat input determined calorimetrically; however, in the
case of the magnetically thin work pieces (2 and 3) this would have created significant
measurement errors. A non-metallic water jacket would be a possible option for cooling without
influencing the electromagnetic properties of the work piece.

Table I. Work pieces and coils used in billet heating experiments

W ork Pieces: 1 2 3 4 5
Alloy: A356 6053 6053 6082 7108
Diameter, mm 76.8 38.6 50.5 95.3 95.4
Length, mm 192 356 1078 258 201/340
Resistivity (ohm m) at 20 C: 4.01E-08 4.11E-08 4.11E-08 3.59E-08 3.92E-08
IACS conductivity, % 43 42 42 48 44
Penetration depth (mm) at 50 Hz: 14.25 14.42 14.42 13.49 14.09
ξw: 3.810 1.893 2.476 4.996 4.788
φ(ξw): 0.819 0.397 0.638 0.852 0.846
ψ(ξw): 1.014 1.060 1.104 1.004 1.004
Coil 1 X X X X X
Coil 2 X X X

Coils: Coil 1 Coil 2


Inside diameter, mm 126 118
Average diameter, mm 132 124
Height, mm 111 300
Coil tube diameter, mm 6 6.35
Coil tube thickness, mm 1.00 0.762
Number of turns 16.5 41
"Aircore" inductance, μH 27.6 71.7
Nagaoka coefficient, Equation (20), kN: 0.653 0.843
Resistance at maximum load, ohms 0.0108 0.0319
Avg. temperature at maximum load, Deg C: 43 62
*
Modified Nagaoka coefficient for load 1, k N : 0.771
*
Modified Nagaoka coefficient for load 2, k N : 0.683
*
Modified Nagaoka coefficient for load 3, kN : 0.704 0.869
*
Modified Nagaoka coefficient for load 4, kN : 0.835 0.935
*
Modified Nagaoka coefficient for load 5, kN : 0.836 0.935
Table II. Electrical measurements, coils 1 and 2 with work pieces 1-5

Load Power
Number V A kVA kW kVAR Factor
1-1 13.14 969 12.7 10.7 7.0 0.84
1-1 14.84 1079 16.0 13.5 8.7 0.84
1-2 13.13 953 12.5 9.8 7.8 0.78
1-2 14.81 1067 15.8 12.5 9.7 0.79
1-3 13.14 954 12.5 10.0 7.6 0.79
1-3 14.81 1067 15.8 12.7 9.5 0.80
1-4 13.13 999 13.1 11.6 6.1 0.88
1-4 14.85 1110 16.5 14.6 7.6 0.89
1-5 13.13 985 13.0 11.4 6.2 0.88
1-5 14.86 1095 16.3 14.4 7.7 0.89
2-3 27.60 703 19.5 16.4 10.5 0.84
2-4 27.58 729 20.1 18.5 7.6 0.93
2-5 27.58 721 19.9 18.5 7.4 0.93

Work piece power has been determined by subtracting the coil power from the power of the
work piece and coil and attributing all resistive changes to the work piece. Where multiple
readings had been taken, averages are shown in Tables II and III. Measured coil power has been
compared to predictions using equations (32) and (42) and against various methods published in
the literature[1, 6, 8]. The methods of Vaughan and Williamson[10] and Tudbury[13], utilizing
the modified Nagaoka coefficient, will give essentially the same results as Equation (42). Short
coil methods of Baker[1] and Burch and Davis[6] and the long coil method of Dwight and
Bagai[8], are shown in Table III for comparison.

Table III. Comparison of real and calculated work piece power by various methods

Load Measured Equation Equation Baker Burch and Dwight and


Number Power, W (32) (42) Short Coil Davis Bagai
1-1 692 754 759 577 821 1275
1-1 899 936 942 716 1013 1583
1-2 138 224 142 125 271 305
1-2 177 281 178 157 340 382
1-3 325 356 318 266 517 641
1-3 377 445 398 333 647 802
1-4 990 1165 1155 862 945 1656
1-4 1264 1438 1426 1064 1167 2045
1-5 1086 1176 1168 866 962 1673
1-5 1422 1454 1443 1070 1188 2067
2-3 640 674 602 569 740 798
2-4 1553 1533 1520 1591 1402 1738
2-5 1923 1810 1797 1627 1657 2054
Error +/-~2% 16.6 6.0 14.6 33.0 67.6
It can be seen in Table III, that Equation (42) produces the best results in comparison with the
measured data, with a typical error of 6% and a maximum error of 16%. Except for work piece
number 2 with ξw<3, Equation (32) gave comparable results with those of Equation (42).

Coil 1 had a length/diameter ratio of 0.84 using the average diameter. While coil 2 had a ratio of
2.42, i.e. coil 2 was relatively long, while coil 1 was quite short. It can be seen in Table III that
the use of the modified Nagaoka coefficient gave an excellent correction for coil shortness, such
that the calculation errors are comparable for both coils using equations (32) and (42). Baker’s
method gives reasonable, but less accurate results for all work pieces and coils. Burch and
Davis’ short coil correction does not work well for small Dw/Dc (work pieces 2 and 3). Dwight
and Bagai’s long coil method is of course only suitable for the “long” coil number 2.

Please note that for coil 2, work piece 4 was shorter than the coil and that the suggested
correction factor of Baker[1] was applied to all methods, i.e. the calculated resistance of the work
piece, e.g. Equation (31), was multiplied by the ratio of the length of the work piece/length of the
coil (lw/lc). This is a direct consequence of Equation (28).

Conclusions

A method has been described using a modified Nagaoka coefficient, capable of estimating work
piece power to better than 10% accuracy, for long and short coils, including cases where the
work piece is shorter than the coil. Tudbury’s work piece shortness factor has been found to be
the square of the modified Nagaoka coefficient as proposed by Vaughan and Williamson in
1945. Given that the presence of a work piece has now been shown to change the magnetic field
inside the air gap of a short coil significantly, the assumption that the impedance of the coil does
not change due to the presence of the work piece can be questioned. It would be of value to
directly measure electrical conditions in the work piece, e.g. circular voltage gradients, in order
to separate the effect of the coil on the work piece and the work piece on the coil.

Acknowledgements

The authors wish to express their gratitude to Egil Torsetnes at NTNU, Trondheim, Norway, for
helping with the design and construction of the experimental apparatus. Deepest gratitude is also
due to Kurt Sandaunet at Sintef, Trondheim, Norway, for the use of the Sintef laboratory and his
contribution in the execution of the experiments. Special thanks to Liss Pedersen at Alcoa, Lista,
Norway, for the supply of filter materials. The author would also like to acknowledge the
funding from the Norwegian Research Council through the RIRA project.
References

1. R. Baker, "Design and calculation of induction heating coils'," AIEE Trans, 57, (1957), 31-40.
2. R. Baker, "Induction Heating of Moving Magnetic Strip," American Institute of Electrical
Engineers, Transactions of the, 64, (1945), 184-189.
3. J. Vaughan and J. Williamson, "Design of Induction Heating Coils for Cylindrical Magnetic
Loads," American Institute of Electrical Engineers, Transactions of the, 65, (1946), 887-892.
4. O. Heaviside, Electrical Papers. (London: Macmillan & Co., 1892), 353-416.
5. C. Burch and N. Davis, "LXX. On the quantitative theory of induction heating," Philosophical
Magazine Series 7, 1, (1926), 768-783.
6. C. Burch and N. Davis, An introduction to the theory of eddy-current heating. (London: E.
Benn Limited, 1928).
7. C. Burch and N. Ryland Davis, "Über eisenlose Induktionsöfen," Electrical Engineering
(Archiv fur Elektrotechnik), 20, (1928), 211-223.
8. H. Dwight and M. Bagai, "Calculations for coreless induction furnaces," American Institute of
Electrical Engineers, Transactions of the, 54, (1935), 312-315.
9. R. Baker, "Heating of Nonmagnetic Electric Conductors by Magnetic Induction---
Longitudinal Flux," American Institute of Electrical Engineers, Transactions of the, 63, (1944),
273-278.
10. J. Vaughan and J. Williamson, "Design of Induction-Heating Coils for Cylindrical
Nonmagnetic Loads," American Institute of Electrical Engineers, Transactions of the, 64,
(1945), 587-592.
11. H. Storm, "Surface Heating by Induction," American Institute of Electrical Engineers,
Transactions of the, 63, (1944), 749-755.
12. N. Stansel, "Induction Heating---Selection of Frequency," American Institute of Electrical
Engineers, Transactions of the, 63, (1944), 755-759.
13. C. Tudbury, Basics of Induction Heating Vol. 1, (New York: John F. Rider, 1960).
14. P. Simpson, Induction heating: coil and system design. (New York, Toronto, London:
McGraw-Hill, 1960).
15. E. J. Davies, Conduction and induction heating. (London: Peter Peregrinus on behalf of the
Institution of Electrical Engineers, 1990).
16. L. Kelvin, Mathematical Papers, Vol. 3,( London: Cambridge University Press, 1890), 484-
515.
17. O. Heaviside, Electrical Papers, Vol. 2, (London: Macmillan & Co., 1892), 396-424.
18. G. Howe, "The high-frequency resistance of wires and coils," Electrical Engineers, Journal
of the Institution of, 58, (1920), 152-162.
19. Copper Wire Tables Circular No. 31: US Bureau of Standards, 1913).
20. V. Callcut, "Coppers for electrical purposes," IEE Proceedings A Physical Science,
Measurement and Instrumentation, Management and Education, Reviews, 133, (1986), 174-201.
21. H. Coppers, "Copper Development Association," Orchard House, Mutton Lane, Potters Bar,
Herts EN6 3AP, (1990).
22. C. Hickman, "Alternating-Current Resistance and Inductance of Single Layer Coils,"
Scientific Papers of the Bureau of Standards, No. 472, (1923), 73-105.
23. S. Butterworth, "Eddy-current losses in cylindrical conductors, with special applications to
the alternating current resistances of short coils," Philosophical Transactions of the Royal
Society of London. Series A, Containing Papers of a Mathematical or Physical Character, 222,
(1922), 57-100.
24. S. Butterworth, "Note on the Alternating Current Resistance of Single Layer Coils,"
Physical Review, 23, (1924), 752-755.
25. S. Butterworth, "On the alternating current resistance of solenoidal coils," Proceedings of
the Royal Society of London. Series A, 107, (1925), 693.
26. A. Arnold, "The resistance of round-wire single-layer inductance coils," Proceedings of the
IEE-Part IV: Institution Monographs, 98, (1951), 94-100.
27. A. Reatti and M. Kazimierczuk, "Comparison of various methods for calculating the AC
resistance ofinductors," IEEE Transactions on Magnetics, 38, (2002), 1512-1518.
28. E. Fraga, et al., "Practical model and calculation of AC resistance of long solenoids," IEEE
Transactions on Magnetics, 34, (1998), 205-212.
29. R. Medhurst, "High Frequency Resistance and Self-Capacitance of Single-Layer Solenoids,"
ed. Wireless Engineer: February, (1947), 35-43.
30. O. Todnem, "Induksjonsoppvarming, Industriell Elektrovarme," ed: NTNU, 1992, 1-60.
31. H. Nagaoka, "The inductance coefficients of solenoids," Journal of the College of Science,
27, (1909), 18-33.
32. E. B. Rosa and F. Grover, "Formulas and Tables for the Calculation of Mutual and Self
Induction," Scientific Papers of the Bureau of Standards, No. 169., (1916), 5-231.
33. H. Wheeler, "Simple inductance formulas for radio coils," Proceedings of the IRE, 16,
(1928), 1398-1400.
34. D. Knight. (2010, August 25). 3.1. Solenoids: Part 1.,
http://www.g3ynh.info/zdocs/magnetics/part_1.html
35. S. Zinn and S. Semiatin, Elements of induction heating: design, control, and applications:
Asm Intl, 1988), 22.